Large Language Models For Automatic Bug Finding In Source Code Analysis H/F - 38

Le CEA est un acteur majeur de la recherche, au service des citoyens, de l'économie et de l'Etat.

Il apporte des solutions concrètes à leurs besoins dans quatre domaines principaux : transition énergétique, transition numérique, technologies pour la médecine du futur, défense et sécurité sur un socle de recherche fondamentale. Le CEA s'engage depuis plus de 75 ans au service de la souveraineté scientifique, technologique et industrielle de la France et de l'Europe pour un présent et un avenir mieux maîtrisés et plus sûrs.

Implanté au coeur des territoires équipés de très grandes infrastructures de recherche, le CEA dispose d'un large éventail de partenaires académiques et industriels en France, en Europe et à l'international.

Les 20 000 collaboratrices et collaborateurs du CEA partagent trois valeurs fondamentales :

- La conscience des responsabilités
- La coopération
- La curiositéJoin us for an internship!
CEA Tech Corporate from CEA Tech on Vimeo
As an intern at CEA, you will have the opportunity to work in a world-renowned research environment. Our teams are made up of passionate and dedicated experts, offering a framework conducive to learning and collaboration. You will have access to state-of-the-art equipment and first-rate research resources to carry out your assignments.

Main goals
(1) Investigate how LLM can be used to assist evaluators in finding bug automatically in source code. For example, a research question is how IA could assist the user in generating formal specification, which is a long repetitive and complex process.
(2) Assess how LLM perform and can be complementary to traditional tools used for evaluation (formal methods, using Frama-C and Lazart).
Internship tasks
-Literature review of LLMs solutions for automatic bug finding.

-Test of LLMs on open benchmarks of source code containing vulnerabilities ([3,4])

-Evaluation of a scope where LLM is relevant (i.e. where it performs better than traditional tools, where it can be complementary, to assist the evaluator)

-Proposition of a methodology to assist source code analysis with LLMs

References
[1] Sauze-Kadar Marine, Thomas. Loubier. (2025). A Multi-Model Approach to Enhance Automatic Matching of Vulnerabilities to Attack Patterns. Récupéré sur https://www.scitepress.org/DigitalLibrary/Link.aspx?doi=10.5220/0013555900003979
[2] Lacombe, G., Feliot, D., Boespflug, E. et al. Combining static analysis and dynamic symbolic execution in a toolchain to detect fault injection vulnerabilities. J Cryptogr Eng 14, 147-164 (2024). https://doi.org/10.1007/s13389-023-00310-8
[3] WooKey challenge: https://wookey-project.github.io/
[4] ANSSI, Amossys, EDSI, LETI, Lexfo, Oppida, Quarkslab, SERMA, Synacktiv, Thales, Trusted Labs. (2020) Inter-CESTI: Methodological and Technical Feedbacks on Hardware Devices Evaluations. https://www.sstic.org/2020/presentation/inter-cesti_methodological_and_technical_feedbacks_on_hardware_devices_evaluations/